Reconfigurable neuron device based on ion gate regulation and preparation method thereof

Abstract

This disclosure provides a reconfigurable neuron device based on ion-gate modulation and its fabrication method, relating to the field of artificial neuron memory device technology. The device comprises, from bottom to top, a synthetic antiferromagnetic layer, a metal oxide layer, an ionic liquid layer, and a top electrode layer. Two opposing antiferromagnetic layers with opposite magnetization directions are disposed on opposite edges of the bottom of the synthetic antiferromagnetic layer. A magnetic tunnel junction for outputting a spike signal is also disposed in the middle of the bottom of the synthetic antiferromagnetic layer. The metal oxide layer, ionic liquid layer, and top electrode layer constitute an ion gate. The ionic liquid layer includes positive and negative ions. When an input voltage is applied to the top electrode layer, oxygen ions in the metal oxide layer move along with the distribution of positive and negative ions in the ionic liquid layer to adjust the charge accumulation at the top interface of the synthetic antiferromagnetic layer, thereby controlling the leakage velocity of the magnetic domain walls at the bottom of the synthetic antiferromagnetic layer through the RKKY effect.

Description

Technical Field

[0001] This disclosure relates to the field of artificial neuron storage device technology, specifically to a reconfigurable neuron device based on ion-gate modulation and its fabrication method. Background Technology

[0002] With social development and technological progress, the read / write speed and power consumption of storage devices are receiving increasing attention. Memory needs high read / write speeds and low operating voltages. To meet these requirements, magnetic memory, due to its non-volatility and fast response speed, is considered an ideal choice for next-generation memory.

[0003] However, current neuronal devices based on magnetic storage structures to simulate biological characteristics suffer from limitations such as the inability to simulate neuronal leakage characteristics, the inability to control leakage characteristics, and difficulty in integration. Neuronal devices implemented using traditional circuits also suffer from drawbacks such as large device footprint and high power consumption. Therefore, current memories designed based on the neuronal characteristics of magnetic storage all exhibit various shortcomings.

[0004] In order to further improve the integration of neuron devices, reduce the power consumption of neuron devices, and better simulate and control the leakage characteristics of neurons, it is imperative to develop an artificial neuron memory device with low circuit design complexity and controllable leakage characteristics. However, existing artificial neuron memory devices mainly have the following defects: (1) They cannot simulate the characteristics of outputting spike signals after continuous accumulation of neurons; (2) They cannot realize the leakage characteristics of neurons; (3) They are not easy to integrate with existing CMOS (Complementary Metal Oxide Semiconductor) technology; (4) The leakage occupies a large area and has high power consumption; (5) The spurious field suppression range is limited and cannot achieve global suppression of neuron devices. Summary of the Invention

[0005] In view of the above problems, this disclosure provides a reconfigurable neuronal device based on ion-gate modulation and its fabrication method.

[0006] The first aspect of this disclosure provides a reconfigurable neuron device based on ion-gate modulation, comprising: a synthetic antiferromagnetic layer, a metal oxide layer, an ionic liquid layer, and a top electrode layer stacked sequentially from bottom to top; a left boundary antiferromagnetic layer and a right boundary antiferromagnetic layer with opposite magnetization directions are disposed on two opposite edges at the bottom end of the synthetic antiferromagnetic layer; a magnetic tunnel junction for outputting a spike signal is also disposed in the middle of the bottom end of the synthetic antiferromagnetic layer; wherein, the metal oxide layer, the ionic liquid layer, and the top electrode layer constitute an ion gate; the ionic liquid layer includes positive ions and negative ions; when an input voltage is applied to the top electrode layer, oxygen ions in the metal oxide layer move along with the distribution of positive and negative ions in the ionic liquid layer to adjust the charge accumulation at the top interface of the synthetic antiferromagnetic layer, thereby modulating the leakage velocity of the magnetic domain walls at the bottom of the synthetic antiferromagnetic layer through RKKY action.

[0007] Furthermore, the top electrode layer is a transparent conductive material, including tin-doped indium oxide; the positive ions and negative ions in the ionic liquid layer are EMI+ ions and TFSI- ions, respectively; the material of the metal oxide layer includes HfZrO.

[0008] Furthermore, the synthesized antiferromagnetic layer comprises a first ferromagnetic layer, a coupling layer, and a ferromagnetic free layer stacked sequentially from top to bottom; wherein, the first ferromagnetic layer is connected to the metal oxide layer, and the first ferromagnetic layer and the ferromagnetic free layer form an antiferromagnetic coupling through the RKKY effect of the coupling layer.

[0009] Furthermore, the first ferromagnetic layer and the ferromagnetic free layer have perpendicular magnetic anisotropy, and their materials each include Co / Pt or CeFeB; the coupling layer is made of at least one of Ru or Ta.

[0010] Furthermore, the magnetic tunnel junction comprises a barrier layer, a ferromagnetic reference layer, and a bottom electrode layer stacked sequentially from top to bottom, wherein the barrier layer is connected to the ferromagnetic free layer.

[0011] Furthermore, the barrier layer is made of Al2O3 or MgO; the ferromagnetic reference layer has perpendicular magnetic anisotropy and is made of Co / Pt or CeFeB; the bottom electrode layer is made of Cu or Au.

[0012] Furthermore, the left boundary antiferromagnetic layer includes a left pinned layer and a left electrode layer stacked sequentially, and the right boundary antiferromagnetic layer includes a right pinned layer and a right electrode layer stacked sequentially, wherein: the left pinned layer and the right pinned layer are each antiferromagnetic and connected to the ferromagnetic free layer, and the magnetization directions of the left pinned layer and the right pinned layer are opposite.

[0013] Furthermore, the materials of the left pinning layer and the right pinning layer each include IrMn, and the materials of the left electrode layer and the right electrode layer each include Cu or Au.

[0014] Furthermore, the magnetization directions of the ferromagnetic reference layer and the first ferromagnetic layer are both negative along the Z-axis; the magnetization directions of the left pinning layer and the right pinning layer are positive and negative along the Z-axis, respectively.

[0015] The second aspect of this disclosure provides a method for fabricating a reconfigurable neuronal device based on ion-gate modulation, characterized by the following steps: growing a left-boundary antiferromagnetic layer and a right-boundary antiferromagnetic layer with opposite magnetization directions at two opposite edges at the bottom of a ferromagnetic free layer; growing an electrode layer at the bottom of the left-boundary antiferromagnetic layer and the right-boundary antiferromagnetic layer; growing a magnetic tunnel junction at the middle of the bottom of the ferromagnetic free layer; sequentially growing a coupling layer and a first ferromagnetic layer on the ferromagnetic free layer, such that the first ferromagnetic layer, the coupling layer, and the ferromagnetic free layer constitute a synthetic antiferromagnetic layer; sequentially growing a metal oxide layer, an ionic liquid layer, and a top electrode layer on the synthetic antiferromagnetic layer, wherein the metal oxide layer, the ionic liquid layer, and the top electrode layer constitute an ion gate, and the magnetic tunnel junction is used to output a spike signal.

[0016] Compared with existing technologies, the ion-gate-controlled reconfigurable neuron device and its fabrication method disclosed herein have at least the following beneficial effects:

[0017] (1) The device disclosed herein can achieve current accumulation, peak signal output, and self-leakage function under the control of RKKY. Under the control of the ion gate, the RKKY is affected, thereby regulating the leakage rate. Thus, the functions of controllable signal accumulation, peak output, and controllable leakage are realized, which relatively completely simulates the function of human brain neurons.

[0018] (2) The device disclosed herein can be combined with a synaptic array based on a magnetic tunnel junction, as well as peripheral circuits such as logic selection circuits and ion gate control circuits to form a neural network circuit and realize certain computing functions. Attached Figure Description

[0019] The above and other objects, features and advantages of this disclosure will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0020] Figure 1 The schematic diagram illustrates the structure of a reconfigurable neuron device based on ion-gate modulation according to an embodiment of the present disclosure;

[0021] Figures 2(a) to 2(b) The diagram schematically illustrates the distribution of oxygen ions controlled by ion gates at different voltages.

[0022] Figure 3 This diagram schematically illustrates the relationship between the position of the magnetic domain wall and time under different RKKY effects during the leakage process;

[0023] Figure 4This diagram schematically illustrates the relationship between the position of the domain wall and time under different RKKY effects during the accumulation process;

[0024] Figure 5 The diagram schematically illustrates an array of ion-gated reconfigurable neuronal devices and a neural network structure according to an embodiment of the present disclosure.

[0025] Figure 6 A schematic flowchart illustrating the fabrication process of an ion-gate-controlled reconfigurable neuronal device according to an embodiment of the present disclosure is shown.

[0026] [Explanation of Labels in the Attached Image]

[0027] 101 - Top electrode layer; 102 - Ionic liquid layer; 103 - Metal oxide layer; 104 - First ferromagnetic layer; 105 - Coupling layer; 106 - Ferromagnetic free layer; BL - Barrier layer; RL - Ferromagnetic reference layer; BE - Bottom electrode layer; 110 - Left pinning layer; 111 - Left electrode layer; 107 - Right pinning layer; 108 - Right electrode layer; Vg - Input voltage; Jc - Current; 201 - EMI+ ion; 202 - TFSI-ion; 203 - Oxygen ion. Detailed Implementation

[0028] The embodiments of the present disclosure will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the disclosure. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the present disclosure for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concepts of the present disclosure.

[0029] It should be understood that when an element (such as a layer, film, region, or substrate) is described as being "on" another element, the element may be directly on the other element, or there may be an intermediate element present. Furthermore, in the specification and claims, when an element is described as being "connected" to another element, the element may be "directly connected" to the other element, or "connected" to the other element via a third element.

[0030] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this disclosure. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0031] Figure 1 A schematic diagram of a reconfigurable neuronal device based on ion-gate modulation according to an embodiment of the present disclosure is shown.

[0032] like Figure 1 As shown, the reconfigurable neuron device based on ion gate control includes: a synthetic antiferromagnetic layer, a metal oxide layer 103, an ionic liquid layer 102, and a top electrode layer 101 stacked sequentially from bottom to top. Two opposing edges at the bottom of the synthetic antiferromagnetic layer are provided with left and right boundary antiferromagnetic layers with opposite magnetization directions. A magnetic tunnel junction 109 for outputting a spike signal is also provided at the center of the bottom of the synthetic antiferromagnetic layer. The metal oxide layer 103, ionic liquid layer 102, and top electrode layer 101 constitute an ion gate. The ionic liquid layer 102 includes positive and negative ions. When an input voltage Vg is applied to the top electrode layer 101, oxygen ions in the metal oxide layer 103 move along with the distribution of positive and negative ions in the ionic liquid layer 102 to adjust the charge accumulation at the top interface of the synthetic antiferromagnetic layer. This, through RKKY (Ruderman-Kittel-Kasuya-Yosida Interaction), regulates the leakage velocity of the magnetic domain walls at the bottom of the synthetic antiferromagnetic layer.

[0033] The top electrode layer 101 is a transparent conductive material, such as indium tin oxide (ITO). Due to its good conductivity and high visible light transmittance, indium tin oxide is used as the top electrode material for ion gates. It is connected to an external circuit, and the input voltage is marked Vg. This voltage Vg is applied to regulate the movement of ions in the ionic liquid layer 102.

[0034] The positive and negative ions in the ionic liquid layer 102 can be, for example, EMI+ ions and TFSI- ions, which can move continuously under voltage regulation.

[0035] The metal oxide layer 103 is made of HfZrO, where oxygen ions, as anions, move to either the lower interface (the interface between the metal oxide layer 103 and the top of the synthesized antiferromagnetic layer) or the upper interface (the interface between the metal oxide layer 103 and the ionic liquid layer 102) as the positive and negative ions in the ionic liquid layer 102 are distributed. It is understood that the proportions of the various elements in HfZrO can be rationally configured to achieve the best oxygen ion regulation effect. It should be noted that the input voltage Vg applied in this embodiment is generally a negative voltage. The magnitude of the negative voltage changes the concentration of oxygen ions at the interface, thereby regulating the strength of the RKKY effect. Therefore, the positive voltage scenario is not shown in the accompanying drawings, but the positive voltage scenario can be derived from the disclosure of this embodiment and falls within the scope of this disclosure, which will not be elaborated further here.

[0036] In this embodiment, the synthesized antiferromagnetic layer includes a first ferromagnetic layer 104, a coupling layer 105, and a ferromagnetic free layer 106 stacked sequentially from top to bottom. The first ferromagnetic layer 104 is connected to the metal oxide layer 103, and the first ferromagnetic layer 104 and the ferromagnetic free layer 106 form an antiferromagnetic coupling through the RKKY effect of the coupling layer 105.

[0037] Specifically, the first ferromagnetic layer 104 and the ferromagnetic free layer 106 have perpendicular magnetic anisotropy, and the materials of the first ferromagnetic layer 104 and the ferromagnetic free layer 106 each include Co / Pt or CeFeB. The material of the coupling layer 105 includes at least one of Ru or Ta.

[0038] In this embodiment, the magnetic tunnel junction 109 includes a barrier layer BL, a ferromagnetic reference layer RL and a bottom electrode layer BE stacked sequentially from top to bottom, wherein the barrier layer BL is connected to the ferromagnetic free layer 106.

[0039] Specifically, the barrier layer BL is made of Al₂O₃ or MgO. The ferromagnetic reference layer RL has perpendicular magnetic anisotropy and is made of Co / Pt or CeFeB. The bottom electrode layer BE is made of Cu or Au.

[0040] Through the embodiments of this disclosure, the device uses ferromagnetic materials with perpendicular magnetic anisotropy, such as CoFeB, as a ferromagnetic reference layer and a ferromagnetic free layer. The ferromagnetic free layer is extended into a structure of a synthetic antiferromagnetic layer. Relying on the RKKY effect of the synthetic antiferromagnetic layer, the movement trend of the domain walls in the ferromagnetic free layer is opposite to the current injection direction, thereby achieving the automatic resetting of the domain walls / neuron leakage function. Simultaneously, by changing the ion gate voltage, the RKKY effect can be modulated, ultimately affecting the regulation of the leakage characteristics of the neuron device.

[0041] In this embodiment, the left boundary antiferromagnetic layer includes a left pinning layer 110 and a left electrode layer 111 stacked sequentially, and the right boundary antiferromagnetic layer includes a right pinning layer 107 and a right electrode layer 108 stacked sequentially. The left pinning layer 110 and the right pinning layer 107 are each antiferromagnetic and connected to the ferromagnetic free layer 106, with the magnetization directions of the left pinning layer 110 and the right pinning layer 107 being opposite. Thus, by growing antiferromagnetic pinning layers on both sides of the bottom end of the ferromagnetic free layer 106, regional pinning of the magnetic domain walls can be achieved.

[0042] Specifically, the materials of the left pinning layer 110 and the right pinning layer 107 each include IrMn, and the materials of the left electrode layer 111 and the right electrode layer 108 each include Cu or Au.

[0043] It should be noted that, Figure 1The diagram schematically illustrates the X-axis, Y-axis, and Z-axis directions. The X-axis and Y-axis directions can be parallel to the top surface of the ferromagnetic free layer 106 and can intersect each other, for example, perpendicularly; the Z-axis direction can be substantially perpendicular to the top surface of the ferromagnetic free layer 106.

[0044] In this embodiment, the magnetization directions of the ferromagnetic reference layer RL and the first ferromagnetic layer 104 are both negative Z-axis, and the magnetization directions of the left pinning layer 110 and the right pinning layer 107 are positive Z-axis and negative Z-axis, respectively.

[0045] Based on this, when no current is injected into the ferromagnetic free layer 106 from the left, the magnetization direction on the left side of the ferromagnetic free layer 106 is negative along the Z-axis, and the magnetization direction on the right side is positive along the Z-axis. Due to the presence of the RKKY effect, the domain walls in the ferromagnetic free layer 106 tend to move to the left, thereby simulating the self-leakage function of a neuron.

[0046] When current Jc is injected from the left side of the ferromagnetic free layer 106, it drives the domain walls in the ferromagnetic free layer 106 to move to the right. When the domain walls move above the magnetic tunnel junction 109, the magnetization direction of the ferromagnetic free layer 106 is reversed to the negative Z-axis, which is consistent with the direction of the ferromagnetic free layer in the MTJ. At this time, the tunneling magnetic resistance of the magnetic tunnel junction is minimal, and a spike signal can be output through an external circuit to simulate the spike output function of a neuron.

[0047] Furthermore, when a negative input voltage Vg is applied to the top electrode layer 101, the positive and negative ions in the ionic liquid layer 102 move upward and downward respectively under the regulation of the input voltage Vg. Due to the charge effect, the oxygen ions in the metal oxide layer 103 move to the lower interface, thereby affecting the RKKY effect of the synthesized antiferromagnetic layer and realizing the functional control of neuronal leakage.

[0048] Figures 2(a) to 2(b) The diagram schematically illustrates the distribution of oxygen ions controlled by ion gates at different voltages.

[0049] As shown in Figure 2(a), when an input voltage Vg is applied to the top electrode layer 101, in the ionic liquid layer 102, EMI+ ions 201 move upward and TFSI- ions 202 move downward. At this time, oxygen ions 203 in the metal oxide layer 103 move downward and are adsorbed onto the interface between the metal oxide layer 103 and the top of the synthesized antiferromagnetic layer.

[0050] Because oxygen ions (203) are adsorbed as negative ions on the top of the synthesized antiferromagnetic layer, i.e., the first ferromagnetic layer 104, a large amount of charge accumulates at the interface, affecting both the first ferromagnetic layer 104 and the ferromagnetic free layer 106. The RKKY coupling between the first ferromagnetic layer 104 and the ferromagnetic free layer 106 in the synthesized antiferromagnetic layer is affected by the accumulated charge. Without an external current, the leftward movement velocity of the ferromagnetic free layer 106 is controlled by the input voltage Vg of the ion gate.

[0051] Furthermore, as shown in Figure 2(b), by adjusting the magnitude of the negative voltage Vg applied to the top electrode layer 101, the accumulation of positive and negative ion charges in the ionic liquid layer 102 at the upper and lower interfaces will be affected, thereby affecting the charge accumulation of oxygen ions at the interface between the ferromagnetic free layer 106 and the metal oxide layer 103. The magnitude of the interface charge will regulate the intensity of the RKKY effect, thereby regulating the leakage speed of the magnetic domain walls at the bottom of the synthesized antiferromagnetic layer through the RKKY effect, and realizing the controllable self-leakage function of the neuron device.

[0052] Figure 3 The diagram illustrates the relationship between the position of the domain wall and time under different RKKY effects during the leakage process.

[0053] like Figure 3 As shown, in this embodiment, the ferromagnetic free layer 106 has a size of 520 nm × 50 nm, and the left pinning layer 110 and the right pinning layer 107 both have a size of 10 nm × 50 nm. As can be seen from the figure, when the RKKY interaction intensity is -0.1 mJ / m... 2 to -0.20mJ / m 2 At this point, the stronger the RKKY effect, the faster the leakage rate of the neuronal device. However, when the RKKY effect is enhanced to -0.25 mJ / m 2 and -0.30mJ / m 2 At that time, there was no significant difference in the self-leakage rate of neurons.

[0054] Overall, the RKKY mechanism enables self-leakage, and the speed of self-leakage in neuronal devices can be controlled by adjusting the magnitude of the ion gate voltage.

[0055] Figure 4 The diagram illustrates the relationship between the position of the domain wall and time under different RKKY effects during the accumulation process.

[0056] like Figure 4 As shown in the figure, in this embodiment, the ferromagnetic free layer 106 has a size of 520nm × 50nm, and the left pinning layer 110 and the right pinning layer 107 both have a size of 10nm × 50nm. As can be seen from the figure, when the input current density is 5 × 10⁻⁶... 7 A / cm2 The relationship between the position of the domain walls and time varies under different RKKY conditions. When the RKKY intensity is low, the domain walls can move to the corresponding positions relatively quickly under the drive of the current; when the RKKY intensity is high, the speed at which the domain walls reach the corresponding positions under the drive of the current decreases.

[0057] In other words, the RKKY effect inhibits the process of the domain wall moving to the right to the excitation signal position, and the accumulation rate of the neuron device can be controlled by adjusting the magnitude of the ion gate voltage and the level of the injection current density.

[0058] Based on the above publicly available information, Figure 5 The diagram schematically illustrates an array of ion-gated reconfigurable neuronal devices and a neural network structure according to an embodiment of the present disclosure.

[0059] like Figure 5 As shown, 501 is a storage array of spin-orbit torque MRAM. For each SOT-MRAM cell, the read word line RWL and write word line WWL control the read and write operations of the SOT-MRAM cell, the bit line BL provides the input, and the source line SL provides the output. When the read word line RWL is on, the SOT-MRAM read operation can be implemented; when the write word line WWL is on, the SOT-MRAM write operation can be implemented.

[0060] 502 is the output array of the ion-gate-controlled reconfigurable neuron device provided in this embodiment. The logic circuit between 501 and 502 is connected to the SOT-MRAM array of 501, and selects a specific neuron device in 502 to output a signal according to its internal logic. In 502, Vin represents the output signal of the logic selection circuit (the input signal of the neuron device), Vcontrol represents the switching control of the logic selection circuit signal, and Vg represents the input voltage applied to the top electrode layer 101 of the neuron device.

[0061] For a single neuron device, the left electrode is connected to the output of the logic circuit; the right electrode is grounded to the lower electrode of the ion gate; the middle MTJ structure outputs a spike signal, represented by the voltage signal Vout. Therefore, the signal output from the SOT-MRAM array is processed in the logic circuit and then output to the neuron device. The operation of the neuron device is controlled by a switch, and the accumulation and self-leakage rates of the neuron device are adjusted by the magnitude of the input signal and the level of the ion gate voltage. This neural network achieves data storage, controllable accumulation of output signals, spike output, and controllable leakage, comprehensively simulating the function of human brain neurons and providing a solid foundation for further in-memory computing.

[0062] Therefore, the neuron device provided in this embodiment has a simple structure and its materials are compatible with CMOS, which is beneficial for large-scale fabrication and integration. This device is capable of realizing neuron accumulation, activation, and self-leakage functions, and has broad prospects in neural computing and artificial intelligence.

[0063] As can be seen from the above description, the embodiments of this disclosure achieve at least the following technical effects:

[0064] (1) The device disclosed herein can achieve current accumulation, peak signal output, and self-leakage function under the control of RKKY. Under the control of the ion gate, the RKKY is affected, thereby regulating the leakage rate. Thus, the functions of controllable signal accumulation, peak output, and controllable leakage are realized, which relatively completely simulates the function of human brain neurons.

[0065] (2) The device disclosed herein can be combined with a synaptic array based on a magnetic tunnel junction, as well as peripheral circuits such as logic selection circuits and ion gate control circuits to form a neural network circuit and realize certain computing functions.

[0066] Based on the same inventive concept, this disclosure also provides a method for fabricating a reconfigurable neuronal device based on ion-gate modulation, which will be described below in conjunction with... Figure 6 The preparation method is described in detail.

[0067] Figure 6 This diagram schematically illustrates a fabrication flowchart of an ion-gated reconfigurable neuronal device according to an embodiment of the present disclosure. The structure of the neuronal device fabricated by this method is shown in the diagram. Figures 1-5 As shown.

[0068] like Figure 6 As shown, the fabrication method of the ion-gate-controlled reconfigurable neuron device may include operations S601 to S604.

[0069] In operation S601, antiferromagnetic layers with opposite magnetization directions, a left boundary antiferromagnetic layer and a right boundary antiferromagnetic layer, are grown at the two opposite edges at the bottom of the ferromagnetic free layer, and an electrode layer is grown at the bottom of the left boundary antiferromagnetic layer and the right boundary antiferromagnetic layer.

[0070] This step achieves pinning of the domain walls at the bottom of the ferromagnetic free layer and injection of current.

[0071] In operation S602, a magnetic tunnel junction is grown at the bottom center of the ferromagnetic free layer.

[0072] This step is used to output the spike signal.

[0073] In operation S603, a coupling layer and a first ferromagnetic layer are grown sequentially on the ferromagnetic free layer, so that the first ferromagnetic layer, the coupling layer and the ferromagnetic free layer constitute a synthetic antiferromagnetic layer.

[0074] This step extends the structure of the ferromagnetic free layer to the structure of the synthetic antiferromagnetic layer. By using the RKKY effect of the synthetic antiferromagnetic layer, the magnetic domain wall movement trend of the ferromagnetic free layer is opposite to the current injection direction, so as to realize the automatic leakage and repositioning function of the neuron.

[0075] In operation S604, a metal oxide layer, an ionic liquid layer, and a top electrode layer are sequentially grown on the synthesized antiferromagnetic layer. The metal oxide layer, ionic liquid layer, and top electrode layer constitute an ion gate, and the magnetic tunnel junction is used to output a spike signal.

[0076] This step involves constructing an ion gate and changing the ion gate voltage to allow oxygen ions in the metal oxide layer to accumulate and adsorb at the lower interface, thereby altering the RKKY effect of the synthesized antiferromagnetic layer and regulating the leakage characteristics of neurons.

[0077] Finally, after fabricating the aforementioned neuronal device, the magnetic domain walls are driven to move to varying degrees depending on the injected current, thereby achieving the neuron's accumulation function. When the magnetic domain walls move to the signal output region, the magnetization direction of the free layer above the reference layer of the output signal magnetic tunnel junction is reversed, aligning the magnetic moment directions of the ferromagnetic reference layer and the ferromagnetic free layer. This results in a spike signal being output through the peripheral circuit, achieving the neuron's activation function. When no current is injected, the magnetic domain walls move in the opposite direction to the current injection direction under the influence of RKKY, achieving the neuron's leakage function.

[0078] It should be noted that the neuron device structure fabricated by the above-described process in the embodiments of this disclosure is as follows: Figures 1-5 As shown, its specific material layers and the structure composed of the material layers are also as follows. Figures 1-5 As shown, further details will not be elaborated here.

[0079] It should be noted that the growth process methods and materials used in the above steps in the embodiments of this disclosure are merely illustrative examples. For instance, the semiconductor layer may also obtain high-quality epitaxial films by using conventional thin film growth and etching methods such as PVD, MBE, ALD, IBE, RIE, ICP, etc. This disclosure does not limit the scope of the invention.

[0080] Furthermore, the above is merely an illustrative example, and this embodiment is not limited thereto. By defining the devices and amplification as described above, and not limited to the various specific structures, shapes, or methods mentioned in the examples, those skilled in the art can make simple modifications and substitutions. For example:

[0081] (1) The size of the neuron device can be appropriately miniaturized or enlarged according to the actual process conditions; the shapes of the electrodes and magnetic tunnel junctions can be modified.

[0082] (2) This ion gate modulates the movement of oxygen ions in hafnium zirconium oxide to achieve charge accumulation and adsorption of the ferromagnetic reference layer by hafnium zirconium oxide. Other ionic liquids can be used to modulate oxygen ions in hafnium zirconium oxide.

[0083] Although the present disclosure has been illustrated and described in detail in the accompanying drawings and the foregoing description, such illustrations and descriptions should be considered illustrative or exemplary rather than limiting.

[0084] Those skilled in the art will understand that the features described in the various embodiments and / or claims of this disclosure can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this disclosure. In particular, the features described in the various embodiments and / or claims of this disclosure can be combined and / or combined in various ways without departing from the spirit and teachings of this disclosure. All such combinations and / or combinations fall within the scope of this disclosure.

[0085] Although this disclosure has been shown and described with reference to specific exemplary embodiments thereof, those skilled in the art will understand that various changes in form and detail may be made to this disclosure without departing from the spirit and scope of the disclosure as defined by the appended claims and their equivalents. Therefore, the scope of this disclosure should not be limited to the above embodiments, but should be defined not only by the appended claims, but also by their equivalents.

Claims

1. A reconfigurable neuron device based on ion gate regulation, characterized in that, The device comprises: a synthetic antiferromagnetic layer, a metal oxide layer (103), an ionic liquid layer (102) and a top electrode layer (101) stacked from bottom to top, opposite edges of the bottom end of the synthetic antiferromagnetic layer are provided with left and right boundary antiferromagnetic layers with opposite magnetization directions, and a magnetic tunnel junction (109) for outputting a spike signal is further provided in the middle of the bottom end of the synthetic antiferromagnetic layer; wherein the material of the metal oxide layer (103) comprises HfZrO, the metal oxide layer (103), the ionic liquid layer (102) and the top electrode layer (101) constitute an ionic gate, the ionic liquid layer (102) comprises positive ions and negative ions, when the top electrode layer (101) applies an input voltage (Vg), the oxygen ions in the metal oxide layer (103) move with the distribution of the positive ions and negative ions in the ionic liquid layer (102) to adjust the charge accumulation at the top interface of the synthetic antiferromagnetic layer, thereby regulating the leakage motion speed of the magnetic domain wall at the bottom of the synthetic antiferromagnetic layer through RKKY effect.

2. The reconfigurable neuron device based on ionic gate regulation according to claim 1, wherein: the top electrode layer (101) is a transparent conductive material, and the transparent conductive material comprises tin-doped indium oxide; the positive ions and negative ions in the ionic liquid layer (102) are EMI+ ions and TFSI- ions respectively.

3. The reconfigurable neuron device based on ion gate regulation according to claim 1, wherein, The synthetic antiferromagnetic layer comprises a first ferromagnetic layer (104), a coupling layer (105) and a ferromagnetic free layer (106) stacked from top to bottom. The first ferromagnetic layer (104) is connected to the metal oxide layer (103), and the first ferromagnetic layer (104) and the ferromagnetic free layer (106) are antiferromagnetically coupled through the RKKY effect of the coupling layer (105).

4. The reconfigurable neuron device based on ion gate regulation according to claim 3, wherein, The first ferromagnetic layer (104) and the ferromagnetic free layer (106) have perpendicular magnetic anisotropy, and each comprises Co / Pt or CeFeB. The material of the coupling layer (105) comprises at least one of Ru or Ta.

5. The reconfigurable neuron device based on ion gate regulation according to claim 3, wherein, The magnetic tunnel junction (109) comprises a barrier layer (BL), a ferromagnetic reference layer (RL) and a bottom electrode layer (BE) stacked from top to bottom, wherein the barrier layer (BL) is connected to the ferromagnetic free layer (106).

6. The reconfigurable neuron device based on ion gate regulation according to claim 5, wherein, The material of the barrier layer (BL) comprises Al2O3 or MgO. The ferromagnetic reference layer (RL) has perpendicular magnetic anisotropy and comprises Co / Pt or CeFeB. The material of the bottom electrode layer (BE) comprises Cu or Au.

7. The reconfigurable neuron device based on ion gate regulation according to claim 5, wherein, The left boundary antiferromagnetic layer comprises a left pinning layer (110) and a left electrode layer (111) stacked in sequence, and the right boundary antiferromagnetic layer comprises a right pinning layer (107) and a right electrode layer (108) stacked in sequence, wherein: The left pinning layer (110) and the right pinning layer (107) each have antiferromagnetic properties and are connected to the ferromagnetic free layer (106), and the magnetization directions of the left pinning layer (110) and the right pinning layer (107) are opposite.

8. The reconfigurable neuron device based on ion gate regulation according to claim 7, wherein, Materials of the left pinning layer (110) and the right pinning layer (107) each include IrMn, and materials of the left electrode layer (111) and the right electrode layer (108) each include Cu or Au.

9. The reconfigurable neuron device based on ion gate regulation according to claim 7, wherein, The ferromagnetic reference layer (RL) and the first ferromagnetic layer (104) each have a magnetization direction of a Z-axis negative direction. The left pinning layer (110) and the right pinning layer (107) have magnetization directions of a Z-axis positive direction and a Z-axis negative direction, respectively.

10. A method of fabricating a reconfigurable neuron device based on ion gate regulation as claimed in claim 1, characterized by, The method comprises the following steps: Growth of a left boundary anti-ferromagnetic layer and a right boundary anti-ferromagnetic layer with opposite magnetization directions on opposite edges of a bottom end of a ferromagnetic free layer, and growth of an electrode layer on a bottom of the left boundary anti-ferromagnetic layer and the right boundary anti-ferromagnetic layer; Growth of a magnetic tunnel junction (109) in a middle of the bottom end of the ferromagnetic free layer; Sequential growth of a coupling layer and a first ferromagnetic layer on the ferromagnetic free layer, so that the first ferromagnetic layer, the coupling layer and the ferromagnetic free layer form a synthetic anti-ferromagnetic layer; Sequential growth of a metal oxide layer, an ionic liquid layer and a top electrode layer on the synthetic anti-ferromagnetic layer, wherein the metal oxide layer, the ionic liquid layer and the top electrode layer form an ionic gate, the magnetic tunnel junction (109) is used to output a spike signal, and a material of the metal oxide layer includes HfZrO.

Images